Stunningly cool photo tech

WOW! Trolling around through links on the web today, I found something very cool.
Microsoft Live Labs (haven’t heard of them before) is working on something called “PhotoSynth”. Video here.
Essentially, they’re doing some massive image feature recognition on popular landmarks. Starting with a huge dataset of photos, they are reconstructing objects in 3D. This is pretty frickin’ amazing stuff. Now, the interface that they have at the moment (Applet here) is not too fabulous, but the idea of taking a huge set of 2D photos from the Net and turning them into navigable 3D spaces is pretty damn compelling. And, perhaps, a potentially cheaper way of capturing real-world environments for use in, I dunno, let’s say, um, games. šŸ™‚
I do wonder whether the photos they’re using now came from particular cameras with known lens characteristics, or if they had some other data in their back pocket, such as position and orientation of the cameras when the photos were taken. If they had none of this data, and are recreating scenes entirely from just the pictures, this is absolutely stunning. (I can see how with a large number of images, they COULD do that, but yowsa!)
Anyway, very cool tech. Check it out!